Picking Winners? Investment Consultants' Recommendations of Fund Managers

7/21/2019 Picking Winners? Investment Consultants' Recommendations of Fund Managers

1/51

Picking Winners? Investment Consultants' Recommendations of Fund

Managers

Journal of Finance, forthcoming

TIM JENKINSON, HOWARD JONES, and JOSE VICENTE MARTINEZ*

ABSTRACT

Investment consultants advise institutional investors on their choice of fund

manager. Focusing on U.S. actively managed equity funds, we analyze the

factors that drive consultants recommendations, what impact these

recommendations have on flows, and how well the recommended funds perform.

We find that investment consultants recommendations of funds are driven

largely by soft factors, rather than the funds past performance, and that their

recommendations have a very significant effect on fund flows. However, we

find no evidence that these recommendations add value, suggesting that the

search for winners, encouraged and guided by investment consultants, is

fruitless.

September 2014

Key words: Asset management, investment consultants, institutional investors, fund performance

JEL classification: G23, G11

! Tim Jenkinson is at the Sad Business School, University of Oxford and CEPR; Howard Jones is at the Sad

Business School, University of Oxford; Jose Martinez is at the University of Connecticut, School of Business. We

are grateful to Andrew Lo, Ludovic Phalippou, Tarun Ramadorai, Peter Tufano and seminar participants at the

Federal Reserve Bank of Cleveland, the University of Connecticut, the University of Western Ontario, the University

of Porto, University of Sussex, ISCTE-IUL, Harvard Law School Pensions and Capital Stewardship conference, the

Seventh Professional Asset Management Conference, the 4th Inquire UK Business School Seminar, and the 2014

Research Affiliates Advisory Panel for valuable comments. We thank Greenwich Associates, eVestment, and

Informa Investment Solutions for making available and explaining their databases, for assistance in building the

combined dataset, and for valuable insights on the analysis. We also gratefully acknowledge research assistance from

Oliver Jones.


2/51

#

Investment consultants are important intermediaries in institutional asset management. Many

retirement plans, foundations, university and other endowments, and other so-called plan

sponsors1engage investment consultants to provide a range of investment services. These include

asset/liability modeling, strategic asset allocation, benchmark selection, active vs passive

management, fund manager selection, and performance monitoring. It has been estimated that, as

at June 2011, almost $25 trillion of institutional assets worldwide were advised on by investment

consultants (Pensions and Investments (2011)). Goyal and Wahal (2008) estimate that 82% of

U.S. public plan sponsors use investment consultants, as do 50% of corporate sponsors.

Furthermore, in some countries plan sponsors are required by law to consult investment

consultants before making their investment decisions.2From the perspective of asset managers,

investment consultants are key gatekeepers, whose opinions determine whether a plan sponsor

will even consider a particular fund. Despite a voluminous literature questioning whether active

managers can add value for investors, many plan sponsors continue to search for active managers.

Investment consultants play a critical role in both encouraging and guiding this search for

winners and so understanding whether they add any value for investors has important

implications for investment strategy.

The investment consulting industry is highly concentrated: measured by assets under

advisement the top ten consultants have a worldwide market share of 82% (and the top ten in the

1We use the term plan sponsors instead of institutional investors to distinguish them clearly from fund managers.

2For instance, U.K. pension fund trustees must obtain and consider the written advice of a person who is reasonably

believed by the trustees to be qualified by his ability in and practical experience of financial matters and to have the

appropriate knowledge and experience of the management of investments (The Occupational Pension Schemes

(Investment) Regulations 2005, regulation 2(2a)).


3/51


4/51

%

around $3 trillion of assets under management.4 Using thirteen years of survey data, we

investigate three questions. First, what drives consultants recommendations? Second, are capital

flows affected by consultants recommendations, i.e. do consultants have substantial influence on

the manager selection decisions of plan sponsors? And, third the main focus of this paper do

these recommendations successfully predict superior performance?

Investment consultants rate products that are aimed at institutional, rather than retail,

investors. A large literature exists on retail mutual funds, and the accuracy of ratings produced by

intermediaries such as Morningstar (Blake and Morey (2000), Khorana and Nelling (1998)).

There is also a recent literature exploring the benefits to retail fund investors of using

professional brokerage firms: Bergstresser et al. (2009) examine these benefits in terms of fund

selection, while Gennaioli et al. (2014) analyze other services of financial advisers, notably the

confidence these firms give to invest in financial assets at all. Much work has also been done on

the accuracy of analyst recommendations for individual stocks (see, for example, Womack

(1996), Barber et al. (2001), Jegadeesh et al. (2004)). On the institutional side, previous authors

have analyzed the performance of investment products (in particular Lakonishok, et al. (1992),

Coggin et al. (1993), Ferson and Khang (2002) and Busse, Goyal and Wahal (2010)) and the

relationship between performance and the hiring and firing of investment management firms

4This is the total at the end of 2011 (the end of our sample period). While the majority of plan sponsors, particularly

the large public pension schemes, employ investment consultants, this figure will include investments in active

equity products from plan sponsors that do not retain an investment consultant. So the total under advisement will

be somewhat less than $3 trillion. We exclude passive index-tracking funds from our analysis, as there is little role

for investment consultants in choosing such products, and they are not included in the recommendations we study.


5/51

&

(Goyal and Wahal (2008)). However, this is the first paper to analyze the formation, impact, and

accuracy of investment consultants recommendations of institutional funds.

The main data we use in this study is provided by Greenwich Associates (GA), which has

conducted surveys of investment consultants since 1988. The U.S. active equity products, on

which we focus, are available for the period 1999-2011. As of 2011, the consultants in the survey

had a 90% share of the consulting market worldwide and 91% of the U.S. consulting market, and

included all of the top ten investment consultants by market share, based on the Pensions &

Investmentsurvey for 2011. The GA survey data tells us, for each year, how many consultants

recommend each fund in a particular size-style category.

We first analyze what drives consultants recommendations of funds by relating

recommendations to the size, fees, and past performance of the funds, and with various non-

performance attributes of fund managers that are evaluated by consultants in the GA surveys.

These non-performance attributes are divided into Soft Investment Factors (i.e. factors which

relate to the investment process) and Service Factors (i.e. factors which relate to service delivery).

We find that consultants recommendations correlate partly with the past performance of fund

managers, but more with non-performance factors, suggesting that consultants recommendations

do not merely represent a return-chasing strategy. We also find that, other things being equal,

larger products attract more recommendations.

Next we compare consultants recommendations of funds with fund flows. We find very

significant flows of funds into, and out of, products following changes in recommendations by

investment consultants; for instance, attracting (or losing) recommendations from one-third of the

investment consultants results, on average, in an increase (decrease) of around 10% or $0.8

billion in the size of the investment product within one year.


6/51

'

Finally we assess the performance of the funds recommended by investment consultants.

To measure fund performance we use standard one-, three- and four-factor pricing models, as

well as returns relative to appropriate size/style benchmarks.We measure performance both gross

and net of fund managers fees.5Starting with returns relative to benchmark, on a value-weighted

basis we find no evidence that recommended products significantly outperform other products.

However, on an equally-weighted basis, we find that average returns of recommended products

are actually around 1% lowerthan those of other products. This result is confirmed using one-,

three- and four-factor pricing models, and the differences using returns against benchmark and

factor models are in every case statistically significant. We also measure the performance of

products in the one- and two-year period after they have experienced a net increase or decrease in

the number of recommendations they receive; we do so in order to test for the possibility that

consultants recommendations add value in the short term, but then become stale and fail to make

a contribution. However, there is no evidence that the net increase or decrease in the number of

recommendations predicts superior or inferior performance respectively.

Given that the more highly-rated funds attract more capital, and larger funds may

themselves underperform owing to diseconomies of scale (as found in the mutual fund sector by

Chen et al. (2004)), we investigate whether the underperformance of the recommended funds

persists having controlled for assets under management. This is because the choice of larger

products may not be a free one: investment consultants may be forced to recommend products of

a certain size, perhaps due to concerns about capacity constraints among small products. We find

a significant negative impact of fund scale on performance, which seems fully to explain the

5Investment consultants themselves charge plan sponsors a fee for their services, which we do not take into account.


7/51

(

underperformance of recommended products. Even controlling for fund size, however, there is no

evidence that the recommendations of investment consultants for these U.S. equity products

enabled investors to outperform their benchmarks or generate alpha.

Our analysis focuses on one asset class, U.S. active equity, which may be more efficient

than other asset classes, and it is possible that elsewhere the recommendations of investment

consultants are more prescient. However, U.S. active equity is a major asset class for plan

sponsors, and our analysis of flows indicates that consultants recommendations in this asset class

are highly influential. This raises the question why plan sponsors engage investment consultants

to help select fund managers without evidence that they add value. We identify three possible

reasons. First, in keeping with the hypothesis of Lakonishok, et al. (1992), plan sponsors may

value the hand-holding service provided by consultants. To use the analogy of Gennaioli et al.

(2014), investment consultants are money doctors with whom investors develop a relationship

of trust, and this in turn gives them confidence when they select fund managers.

Second, investment consultants may provide a shield that plan sponsors can use to defend

their decisions. This is in keeping with the finding of Goyal and Wahal (2008) that plan sponsors

most sensitive to headline risk are most likely to use investment consultants, and also with the

conclusions of Jones and Martinez (2014) that plan sponsors disregard their own expectations of

fund managers performance in favor of the recommendations of investment consultants.

Third, as a result of consultants lack of transparency and their own naivety, plan sponsors

may misunderstand the utility of these recommendations (cf. Inderst and Ottaviani, (2012) for a

perspective on retail financial advice in this vein). While consultants insist on full transparency in

the performance of the fund managers they rate, they do not themselves disclose past

recommendations to allow evaluation of their own performance. Our industry-wide analysis

shows that non-recommended funds perform at least as well as recommended funds. Of course,


8/51

)

some consultants will be more accurate than others, but plan sponsors do not have sufficient

information to tell them apart. In the light of our findings, a natural response by plan sponsors (or

regulators) would be to require investment consultants to provide the same level of disclosure as

that which is provided by fund managers on their performance, or the same level of disclosure

provided by research analysts on their stock recommendations.6

The remainder of this paper is structured as follows. In the next section we describe the

role of investment consultants in selecting institutional funds. In section 2 we describe our data

set. In section 3 we describe our methodology and we analyze the drivers of consultants

recommendations, the capital flows that they bring about, and the performance difference

between recommended and non-recommended products. We also conduct robustness checks on

our main results. Section 4 concludes.

1. The Role of Investment Consultants

Investment consultants provide several types of advice to plan sponsors.7 In the sequence of

decision-making, the first engagement, which is carried out when a plan is formed and

periodically thereafter, is to help the sponsor determine and formulate the objectives of the plan.

These vary significantly between plan types (e.g. pension funds and endowments). Next, the

6It is market practice for U.S. institutional fund managers to comply with the requirements of the Global Investment

Performance Standards (GIPS) when disclosing their performance. As for research analysts, FINRA Rule 2711(h)(5)

requires brokers to disclose the aggregate percentage of Buy, Hold and Sell recommendations for the set of all the

companies they cover for the previous twelve months. However, market practice is for brokers also to disclose their

actual recommendations of individual stocks for the previous three to five years and, in some cases, even longer.

7For a description of the role of investment consultants see SEC (2005).


9/51

*

consultant advises the plan sponsor on an investment strategy to accomplish those objectives,

including the choice of benchmarks, the broad allocation of assets between asset classes, and

agreed bands within which this allocation may vary. Within some asset classes the next decision

is whether to opt for an active or a passive approach and then to select fund managers under the

approach adopted. After the appointment of fund managers, the consultant monitors their

performance, and may make recommendations for termination and replacement.

Why do plan sponsors employ investment consultants for manager selection? In most

cases, ultimate fiduciary responsibility for the performance of the assets rests with trustees who

are non-specialists and require independent and specialist advice. Day-to-day management of the

assets is typically carried out by investment professionals employed by the plan, but the trustees

are ultimately responsible for hiring, monitoring and firing the investment professionals and fund

managers employed by the plan, as well as strategic asset allocation decisions. Investment

consultants report directly to the trustees and provide them with information and advice to allow

them to discharge their responsibilities.

The scope of the advice sought from investment consultants depends on the professional

skills of the trustees and the extent of in-house expertise, as well as the complexity of the

investment strategy being followed. An index-tracking strategy for equity investments requires

limited input from consultants, as the choice between products is relatively simple and passive

managers require limited monitoring. However, in the case of active managers, investment

consultants are asked to make recommendations for both hiring and firing drawing on their

program of fund manager due diligence. This research involves both quantitative analysis of past

performance and qualitative judgments about the fund manager. We discuss the various

qualitative factors in the next section.


10/51

+

Plan sponsors use consultants recommendations both when they first hire and when they

replace managers. As part of the hiring procedure, the consultant typically draws up a shortlist of

its recommended fund managers. In some cases the plan sponsor, for fiduciary reasons, makes

the final selection alone; in other cases the plan sponsor takes further advice from the investment

consultant when choosing from the shortlist. These consultant services are paid for by a retainer

if they are recurrent (e.g. manager monitoring) or under a schedule of charges for ad hocwork

(e.g. manager termination and selection).8

The importance of investment consultants to plan sponsors is reflected in survey data. The

Pensions and Investments (2011) survey of plan sponsors found that 23% of respondents rated

investment consultants as crucial to the operation and success of their plans, with a further 40%

rating consultants as very important, and 26% rating them somewhat important. As noted

earlier, investment consultants offer a range of services, but when plan sponsors were asked in

what area they felt their consultants added the most value, the most frequent responses were

money manager search/selection (27%), asset allocation development (27%) and

performance measurement/reporting (23%). Similarly, on the sell-side, investment

management firms view investment consultants as thekey gatekeeper to plan sponsors.

Investment consultants do not disclose past recommendations in a way that would allow

their accuracy to be measured. Some consultants show their value added by comparing the

performance of a portfolio of their recommended funds with that of a chosen benchmark.

However, they do not generally compare this performance with the performance of institutional

8It is worth noting that whereas recommendations of retail funds are often provided free of charge, or, in the case of

equity analysts, such research is bundled together with brokerage services, the advice of investment consultants on

fund managers is paid for directly by the plan sponsors.


11/51

#,

funds which they do notrecommend, nor do they make available the underlying data for scrutiny

by third parties. To overcome this constraint, we use the leading industry survey of investment

consultant recommendations, which we describe in the next section.

2. Data

Our main source of data is a survey of investment consultants recommendations of U.S.

long-only (i.e. excluding hedge funds) actively managed equity products that Greenwich

Associates has been conducting annually since 1988. In this survey investment consultants are

asked to rate active fund managers on various measures of performance and service, and also to

state the names of the fund managers they recommend to their clients for each of a number of

investment styles. Further details on the GA surveys can be found in the Internet Appendix.

We draw on the surveys between 1999 and 2011. For the period before 1999 the GA

survey does not contain information on investment consultants recommended products. Each

year the survey was carried out over a two-to-four month period starting between late November

of one year and early January of the next. Consultants respond to the questionnaires in confidence,

and the responses by individual investment consultants to the GA questionnaires are not

disclosed in the survey results, but rather the aggregate responses.

The main information we obtain from these surveys is an annual list of fund managers

showing, in each size/style category, the percentage of the consultants surveyed who

recommended that fund manager.9According to GA, consultants are asked to recommend

9Many investment consultants will, having performed their due diligence on a fund manager, assign a rating similar

to the buy-hold-sell classification used for equity analysts. However, the survey data we use focuses on their positive


12/51

##

between four and six fund managers for each of seven different market-capitalization-style sub-

categories: Large Cap Growth, Large Cap Value, Small Cap Growth, Small Cap Value, Mid Cap

Growth, Mid Cap Value and Domestic Equity Core. If a fund manager manages more than one

product in a given size/style category we aggregate those products into a single one, to make it

correspond to the GA classification.10Since we obtain our data from original documents we are

confident that all recommendations are included in the database even if a product ceases to exist,

or if returns are no longer reported, and so the recommendations data are free from survivorship

and backfill bias.

In addition to providing a shortlist of recommended products, consultants responding to

the survey express their opinions on a fund manager in an entire asset class rather than on

individual products or groups of similar products within that class. GA divides the headings

under which respondents are asked to rate fund managers into two sets: investment factors and

service factors. Three of the investment factors, which we call soft investment factors, are:

clear decision making, capable portfolio management, and consistent investment philosophy. The

service factor category includes the capabilities of relationship professionals, usefulness of

reports prepared by the fund manager, effective presentations to consultants, as well as some

other factors that varyfrom year to year. For each factor the respondent is invited to rate a fund

managers performance in each asset class on a five-point scale. Under each factor GA then

aggregates the responses into a single score for each fund manager in each asset class surveyed

recommendations. Consequently, in the empirical work we compare the performance of the recommended (buy)

funds to all others (hold, sell and those that have not been analyzed).

10This was not very common: only 19% of the observations in our sample represent aggregated products in the same

size/style category.


13/51

#$

using the Rasch model (see Andrich (1978)). In this study we work with a modified version of

these scores, the fractional rank, obtained by ranking the fund managers scores for each variable

into percentiles and dividing them by 100 to arrive at a factor for each manager between zero and

one.

We combine this GA data on investment consultants with data for the same period on the

returns of institutional U.S. equity funds, their asset management fees, and their assets under

management, all provided by eVestment, and with additional fee data from Informa Investment

Solutions (IIS).11The eVestment databases report the monthly returns of institutional funds, their

assets under management (at an annual, and sometimes quarterly, frequency), and the latest

available fees charged by those funds. From IIS we obtain year-by-year asset management fees,

which are not available from eVestment. The returns we obtain for the products in the eVestment

database are composite returns. The individual returns earned by each client may deviate from

these composite returns, but deviations are typically small.12Composite returns are net of trading

costs, but gross of investment management fees. The data are self-reported by the fund managers,

but constant scrutiny from clients using this data guarantees a high degree of accuracy. The return

data are free from survivorship bias: like the GA survey, the eVestment database retains data for

funds that have been discontinued (e.g. because they have been acquired or closed). For each

product, the databases also provide cross-sectional information (as of June 2012) on investment

style and capitalization bracket, manager-designated benchmark, and the latest fees.

11eVestment and IIS are both leading providers of data and analytic services to institutional fund managers.

12 For example, some investors may require that their part of the overall portfolio is purged of the influence of

companies involved in arms, tobacco, alcohol, gambling etc.


14/51

#%

To match the eVestment data with the GA asset class of U.S. long-only equity, from

where our sample of shortlisted products is drawn, we first eliminate index funds (including

enhanced index funds), hedge funds, REITs and retail funds. We also eliminate products that do

not match the seven GA size/style sub-categories.13 We also drop observations for Mid Cap

Growth and Value products before 2001 and for Equity Core products before 2003, as the GA

survey did not ask for recommendations on these products before these dates.

Tables I and II provide descriptive statistics for the final sample. As Table I shows, the

average number of available products during the sample period is 1,919. Approximately 21% of

the products received at least one recommendation each year from an average of 29 investment

consultants who answered the survey each year. Average assets per product are $1.5 billion.

Recommended products are substantially larger than non-recommended ones: average assets

managed by recommended product are $4 billion, whereas the average size of non-recommended

products is only $0.8 billion. At the end of our sample period the total assets under management

of (recommended and non-recommended) products in the sample are $3 trillion. This figure is

similar to that reported by Busse et al. (2010) for 2008, which suggests that the coverage of the

databases is similarly comprehensive.14

13We match U.S. Equity Core products in the GA surveys to Large Cap Core, Mid-Large Cap Core, Mid Cap Core,

and All Cap Core products in the eVestment databases.

14$3 trillion is the total value, at the end of 2011, of U.S. active equities managed by institutional asset managers in

our sample, having excluded categories that do not match the GA survey data, including around $1.2 trillion of U.S.

passive products.


15/51

#&

Table II shows average annual pro forma fees (in percent) of the recommended and non-

recommended products.15Panel A shows fees using the eVestment database as of the last year of

the sample, 2011, or latest available. Fee information for each product is not available from

eVestment for each year. However, to check whether there have been significant changes in fees

we also source data from IIS, which is available on an annual basis. Panel B shows average fees

in the IIS database over the 13-year period covered by our study (1999 to 2011). As is typical in

this industry, fees decline as investment levels increase. There is, however, no difference between

average fees charged on recommended and not recommended products. For a $50 million

mandate the average fee of recommended products is 68 bps compared with 69 bps for non-

recommended products. There are no significant differences between using eVestment or IIS as

the source of data. We also find very little time series variation in fees, and we do not find much

intra-style cross sectional variation (see the Internet Appendix for details). These findings are

very much in line with Busse et al. (2010). We also find that, in contrast to the strong correlation

between fees and account size, there is (almost) no relationship between fee and product size

(product assets under management), as the correlation between these two variables is on average

just -0.06 (average across categories) and is never statistically significant.

3. Methodology and Results

The first part of this section investigates the factors that influence investment consultants

recommendations of U.S. actively managed equity products. The factors we analyze include not

15Pro forma fees are not necessarily the same as actual fees, which are the result of a confidential negotiation process.

However, we understand from industry sources that fee negotiations are typically conducted between plan sponsors

and fund managers and occur afterthe investment consultant has drawn up the shortlist of recommended products.


16/51

#'

only the past performance of the product, but also the non-performance attributes which the

consultant identifies in the fund manager the Soft Investment Factors and Service Factors as

well as product size and fees. In the second part we assess the impact of investment consultants

recommendations on flows into and out of these investment products, showing the extent to

which investment consultants recommendations are actually followed by plan sponsors. In the

third part we examine whether investment consultants recommendations of fund managers add

value to plan sponsors, by comparing the performance of recommended and non-recommended

products. In the fourth part we test the robustness of our performance findings and conduct

further analysis, notably to investigate whether differences in size between recommended and

non-recommended products can explain the observed underperformance of the former.

A. Drivers of Recommendations

In this section we explore the determinants of investment consultants recommendation

decisions. We estimate a Poisson model, with the standard exponential mean parameterization, on

pooled yearly data. The pooled Poisson estimator assumes that the number of recommendations

received by a product i at time t(Recsi,t) is Poisson distributed with a mean of:

! !"#$!!! !!!! ! !"# !! ! !!!"#$!"#$!!! ! !!!"#$!"#!!"#$%&'!!! ! !!!"#$%&"!"#$%&'!!! !

!!!""!!! ! !!!"#!!!!! ! !!!"#$%&!"#!!! (1)

Past Perfi,tis either the gross return or Fama-French three factor alpha fractional rank of product i

in relation to the other products in the same market capitalization and style category over the two

year period immediately preceding the survey from which recommendations are collected. Soft

Inv. Factorsi,tis the fractional rank at time tof a set of Soft Investment Factors of fund manager

i's U.S. equity team (i.e., Clear Decision Making, Capable Portfolio Manager, and Consistent


17/51

#(

Investment Philosophy). Service Factorsi,t is the fractional rank at time t of a set of Service

Factors of fund manager i's U.S. equity team (i.e., Capabilities of Relationship Professionals,

Usefulness of Reports prepared by the fund manager, Effective Presentations to consultants).

Feei,Tis the fractional rank at time T(i.e. the end of the sample period) of the asset management

fee of product ibased on a $50m investment in relation to the other products in the same market

capitalization and style category. AUMi,t-1is product isassets under management (in $ billions)

at time t-1. Finally, Return Voli,t is a measure of product is return volatility over the two year

period preceding the survey.

When estimating this model we use robust standard errors clustered at the product level

and a full set of time dummies. The Poisson Quasi-MLE estimator we use retains consistency if

the count is not actually Poisson distributed, provided that the conditional mean function is

correctly specified (see Wooldridge, (2010)).

Table III shows the results of this exercise. The table contains both coefficient estimates

and average marginal effects. Models I and II use the soft investment and service quality indexes

as regressors whereas Models III and IV replace them with (some of) their components.16Results

suggest that investment consultant recommendations are at least partly driven by past good

performance (acommon phenomenon in financial analysts decisions to recommend stocks; see

Altinkilic and Hansen, (2009)). Moving from the bottom to the top percentile of past performance

leads, on average, to an increase of half a recommendation in the average number of

recommendations received by a product (out of an average maximum of 29 recommendations per

16We exclude Service Factors that were part of the survey for only part of the sample period.


18/51

#)

year, the average number of consultants issuing recommendations in our sample). 17 Soft

Investment Factors, notably CapablePortfolio Manager and Consistent Investment Philosophy,

seem to have a more important impact on recommendations. Estimates in Models I and II

indicate that moving from the bottom to the top of the Soft Investment Factors ranking is

rewarded, on average, with slightly under six extra recommendations per year. Service Factors,

and in particular Capabilities of Relationship Professionals and Usefulness of Reports, also

appear to be important drivers of recommendations. Estimates in Models I and II suggest that

improvements in service quality, from the bottom percentile to the top one, lead to one extra

recommendation received per product per year.

Larger products are associated with a higher number of recommendations, perhaps

reflecting the concentration of the investment consulting sector, with consultants focusing on

products that are suitable for their range, and scale, of mandates. In these regressions lagged

assets under management is a predetermined variable and therefore contemporaneously

exogenous. As a robustness check, and to account for the possibility that current

recommendations might affect (one period) lagged AUM on the grounds that current

recommendations might have been issued a number of months before they appear in the survey

we also estimate an instrumental variable version of this model. In the Internet Appendix we

report the result of using the GMM estimator of Mullahy (1997) to address the potential

endogeneity ofAUMi,t-1usingAUMi,t-2as an instrument.

17The time period and performance models we use to gauge past performance could differ from the consultants own

assessment of past performance. Any such a difference could serve to attenuate the estimated impact of past

performance on consultants recommendations. It is however reassuring that, as shown in the Internet Appendix,

including additional past performance measures in our model does not significantly affect our results.


19/51

#*

One surprising result is that, if anything, higher fees are associated with an increased

probability of recommendation. However, given the tight clustering of fees observed in Table II

this effect is not economically significant. Moreover, as noted earlier, since plan sponsors

typically negotiate fees with fund managers after the investment consultants have drawn up a

shortlist of candidates, actual fees paid by plan sponsors are likely to show more variation than in

the eVestment and IIS data, and therefore these results should therefore be treated with caution. 18

Although an analysis of the distribution of the number of recommendations per product

indicates that the data is over-dispersed, thus violating the Poisson variance assumption, the

Quasi-MLE estimator used to estimate Models I to IV is robust to this problem. For further

robustness, however, Models V to VIII in Table III repeat the estimation using the Negative

Binomial (NB2) of Cameron and Trivedi (1986), a particular parameterization of the negative

binomial distribution that allows for over-dispersion, instead of the Poisson distribution. As

shown in columns V to VIII, the results are nearly identical to those of the Poisson model,

suggesting that both models provide similar fit for the conditional mean.

To summarize the results in this section, we find that investment consultants

recommendations are a function of past fund performance, but especially of the two sets of non-

performance factors (Soft Investment Factors and Service Factors) that consultants identify in

fund managers. In particular, Soft Investment Factors appear to have a far more powerful effect

on consultants recommendations than past performance. So although consultants

recommendations to some extent reflect a return-chasing strategy, they seem to be more heavily

18In the Internet Appendix we explore alternative fee specifications, including using time t-1 and time-averaged fees

from IIS, finding no significant differences in results.


20/51

#+

influenced by the consultants qualitative judgment of intangible factors. We also find that, other

things being equal, larger products attract more recommendations.

B. Flows

In this section we explore how asset flows respond to changes in consultants

recommendations.19 We do this by expanding a typical flow-performance regression (see, for

instance, Ippolito (1992), Chevalier and Ellison (1997) and Sirri and Tufano (1998)) to include a

recommendation change variable as regressor. We consider two flow measures. The Dollar Flow

into and out of an investment product is defined as the yearly change in the total net assets minus

appreciation:

!!"#$!!! ! !"#!!! ! !"#!!!!! ! !! !!!! (2)

where TNAi,tis the total net assets for product iat date t, and ri,tis the gross return on product i

between dates t-1andt. This measure reflects the growth of a fund in excess of the growth that

would have occurred if no new funds had flowed in but dividends had been reinvested.

The second measure is the Percentage Flow relative to the total net assets invested in the

product as of the end of the previous year:

!!"#$!!! !!!"#$!!!

!"#!!!!! (3)

19 Jones and Martinez (2014) show that plan sponsors follow investment consultants recommendations of fund

products. Their analysis focuses on the incremental effect on flows of consultants recommendations, over and

above the effect of such recommendations that is channelled through the other variables of interest, notably through

the expectations of plan sponsors themselves. In this paper we measure the full effect of changes in

recommendations.


21/51

$,

The bivariate relationship between recommendation changes and Dollar and Percentage Flows is

shown in Figure 1.20The graph plots show the results of estimating kernel weighted local linear

regressions of Dollar Flows (Panel A) and Percentage Flows (Panel B) on lagged changes in

consultants recommendations. The results indicate a positive relationship between the change in

consultants recommendations (measured as the change in the percentage of recommendations

received by a product over the total possible) and subsequent flows. We are interested, however,

in measuring how flows respond to recommendation changes, controlling for publicly-available

measures of past performance, as well as for other product attributes known to affect flows and

which could also affect recommendations (namely past performance, fund size, and return

volatility). We estimate the response of flows to recommendation changes using the following

regression on yearly data:

!"#$!!! ! !! ! !!!"#$%&'()$(%!"#$!!!!! ! !!!"#$!"#$!!!!! ! !!!"#$%"&'!!!!! ! !!!!(4)

Flowi,tis either the Dollar Flow or the Percentage Flow. !"#$%&'()$(%!"#$i,t-1is the change in

the number of recommendations product ireceived, as a fraction of the highest possible number

of recommendations which that product could have received from all of the consultants in our

sample, between time t-2and t-1. Past Perfi,t-1 is a set of performance measures of product iat

date t-1(the one year gross return and Fama-French three factor alpha rankings of a product in

relation to the other products in the same market capitalization and style category). The

regressions also include the total net assets for product i at date t-1 (in the Percentage Flow

20To reduce the effect of outliers on the coefficient estimates, we winsorize the percentage flows variable at the 95th

percentile (as in Barber et al. (2005)).Similar results obtain if we remove small products from the sample instead.


22/51

$#

regression we use log assets as the regressors; see Del Guercio and Tkac (2002)), a measure of

product return volatility over the previous two years, and a full set of time dummies (one per each

year in the sample) as additional controls.

Table IV reports the results of estimating this regression using pooled time-series cross-

sectional data with Dollar Flows and Percentage Flows as the dependent variables. Each column

in this table represents a separate regression. The coefficients of the variables capturing the effect

of lagged changes in recommendations on absolute flows are positive and statistically significant.

This suggests that plan sponsors respond to the investment consultant recommendation changes

by moving money in the direction implied by the recommendation change. The estimate in

column I indicates that moving from a situation where no consultant recommends the product to

another when all of the consultants in our sample recommend it leads to an extra inflow of assets

of $2.4 billion.

Qualitatively similar results obtain if we use percentage flows as the dependent variable.

Estimates in column IV suggest that a product shortlisted by all the consultants in the sample, in

the previous year, receives, on average, extra inflows equal to 29% of the assets managed by that

product in the previous year, compared to a similar product that is not shortlisted by any

consultant. In all cases t-statistics are based on clustered standard errors, which are White

heteroskedastic-consistent standard errors corrected for possible correlation across observations

of a given investment product (White (1980) and Rogers (1993)). This method seems to be the

most appropriate given the size of our panel (see Petersen (2009)).

The difference between, on the one hand, columns I and IV and, on the other, columns II

and V is that II and V include also lagged recommendation levels (as opposed to changes) as

regressors while I and IV do not. The economic reason for including lagged recommendation

levels is that institutional money may be slow to react. This regressor reflects the extent to which


23/51


24/51

$%

affects assets flows is the products gross return ranking in relation to the other products in the

same category and not its risk-adjusted return ranking. This is consistent with the widespread use

in the industry of this measure.

To summarize, we find that changes in investment consultants recommendations have a

large and significant effect on flows into institutional investment products.

C. Performance

We measure the performance of consultants recommended products by estimating factor

models using time-series regressions. To generate aggregate measures of performance, we create

equal- and value-weighted portfolio returns of recommended and not recommended products

available in each month. In this analysis the recommended portfolio includes each fund as many

times as it is recommended. The weight used for value-weighting is based on the assets in each

product at the end of December of the prior year. With these returns, we estimate:

!!!! ! !! ! !!!!! ! !!!! (5)

where rp is the portfolio excess return (over the risk-free return), ft is a vector of excess returns

on benchmark factors, and !p is the abnormal performance measure of interest. We use three

established factor models: CAPM (Sharpe (1964), Jensen (1968)), the Fama-French (1993) three

factor model and the Fama-French-Carhart four factor model (Carhart (1997)). We obtain these

four factors from Kenneth Frenchs web site.22In addition to these measures we report the

22Academic factor models are more demanding than practitioner benchmarks, and fund sponsors may give credit to

fund managers for allocation decisions that are not reflected in alpha under such models. For example, although

practitioners use style benchmarks such as small caps, investing in very small stocks could enhance a funds


25/51

$&

average returns of the products in excess of a selected benchmark. The benchmarks we use are

listed in the Internet Appendix and are standard in the investment industry.

We work with two versions of these performance measures: one based on gross returns

and another one based on net returns. To compute net returns-based measures, we subtract one-

twelfth of the annual pro forma fee based on a $50 million investment from the products

monthly return, using the fee information from eVestment summarized in Table II.23

Results in Table V indicate that, in the 13-year period of our study, and on an equal-

weighted basis, the portfolio of all products recommended by investment consultants delivered

average returns net of management fees of 6.31% per year (7.13% before management fees).

These returns are, on average, 1.12% per annum lower than the returns obtained by other

products available to plan sponsors, which are not recommended by consultants. When we risk-

adjust returns using the three- and four-factor models, recommended products obtain an alpha of

performance against this benchmark, but this outperformance would be factored out in academic models. This is

not the only source of differences between the approaches. Cremers et al. (2013) note that the practitioner and

academic approaches can yield very different results, as the standard Fama-French and Carhart factor models may

assign non-zero alphas even to passive benchmark indices such as the S&P 500 and Russell 2000. Also, the

classification system on which benchmark-adjusted returns are based is ambiguous and leaves room for

interpretation resulting in frequent misclassifications (some perhaps deliberate). For example, in a study of U.S.

mutual funds, diBartolomeo and Witkowski (1997) find that almost 40% of all U.S. equity funds are misclassified.

23We focus in the remainder of the paper on net returns using fees obtained from eVestment for the last year of our

sample (or latest available year for the product), since eVestment has a more complete coverage of our sample of

products than the alternative source of fee data, IIS (which reports fees on a year-by-year basis). We estimate several

models using IIS data in the Internet Appendix with no change to the conclusions. Note that none of our results take

into account the impact of the fees of the investment consultants themselves.


26/51

$'

0.39% per year, once annualized. This is still significantly lower than the alpha obtained by non-

recommended products (the difference between recommended and non-recommended products is

statistically significant at -0.97% per year). Risk-adjusting returns using benchmarks chosen to

match the products style and market capitalization delivers almost identical results. When we

run similar regressions among the recommended products, separating the most recommended

from the least recommended, the results also show that the underperformance of the

recommended products on an equal-weighted basis is mostly concentrated among the most

frequently recommended products (for details see the Internet Appendix).

When we perform the same analysis on a value-weighted basis recommended products

still obtain lower returns (or CAPM alphas) than those obtained by non-recommended products,

but outperform them based on a three- or four-factor model. None of these differences is

statistically significant, however. Value-weighted returns and alphas are consistently lower,

suggesting that smaller products perform relatively better.24

In Table VI we separately consider the subcategories within our sample: Large Cap Value,

Large Cap Growth, Mid Cap Value, Mid Cap Growth, Small Cap Value, Small Cap Growth and

Domestic Equity Core products. The results we obtain broadly confirm those presented in Table

V: recommended products underperform other products in all the categories studied when returns

are equal-weighted. On a value-weighted basis results are mixed.

As noted earlier, the investment consulting industry is highly concentrated with the top 10

consultants having 81 percent of the US market. The recommendations issued by the smallest

24Our equal- and value-weighted figures, once aggregated across recommendation categories, are generally in line

with those reported by Busse et al. (2010) for their sample of institutional products.


27/51

$(

half of consultants may therefore be less relevant in terms of impact, while still receiving the

same weight as the recommendations issued by the largest consultants in our analysis. To address

this concern we explore whether there are any substantial differences between recommendations

issued by large and small investment consultants (for a subset of the years for which GA had

separate data available for both groups of consultants). As reported in the Internet Appendix, we

find little difference between the recommendations of large and small consultants: their

recommendations seem to be highly correlated, and average product size, average asset

management fees and past performance rankings of recommended products are similar.

Our results so far suggest that investment consultants are not able consistently to add

value by selecting superior investment products. This is particularly true when we compare

recommended products with non-recommended products on an equal-weighted basis.

D. Performance Robustness and Further Analysis

How can we explain these findings, given that we should expect recommended products

to outperform other products by a margin sufficient to cover the cost of hiring the investment

consultant? Why do the investment consultants select products that appear to underperform other

products significantly on an equal-weighted basis, but not on a value-weighted basis? In this

section we explore these issues in more detail. First we explore the possibility that backfill bias

may affect our results and, in particular, that it may be responsible for the relatively good

performance of non-recommended products, which would therefore constitute an unfair

benchmark for recommended products. Second, we examine the impact of investment product

size on performance. This allows us to assess the relative performance of recommended and the

(generally) smaller non-recommended products controlling for the potential effect of product size

on performance. Finally, we investigate the short-term performance of products that experience a


28/51

$)

net change in the number of recommendations they receive, as it is possible that consultants

failure to add value is due not to their inability to identify outperforming products, but to the fact

that they continue to recommend those products for too long.

Because we work with self-reported returns a natural concern in the measurement of

performance is that the data might not accurately reflect the performance of worse performing

products. Since managers may have a greater incentive to volunteer information to eVestment

after a period of good performance, products may be subject to instant history or backfill bias,

as described by Fung and Hsieh (2000). Moreover, this problem could be more serious for

smaller non-recommended products if they are subject to less scrutiny than their recommended

counterparts. It is unlikely that this problem can ever be completely eliminated, but we follow the

approach in Jagannathan et al. (2010) to determine the potential impact on our results. We

eliminate the first one, two and three years of returns for each product and re-run our main

performance regressions. By requiring three years of return history to show fund performance,

this procedure also addresses another concern with our performance comparison: most pension

sponsors and consultants require the existence of a three-year performance track record to be

considered in the initial phases of a manager search (Del Guercio and Tkac, 2002).

Results reported in Table VII suggest that performance figures may have a backfill bias,

but the evidence we find suggests that it is not enough to affect our main results. After

eliminating the first three years of return history, on an equal-weighted basis the three-factor

alphas of non-recommended products decline by only 0.25% (from 1.36% to 1.11%), and those

of recommended products decline by 0.17% (0.39% to 0.22%). The difference is significantly

smaller than the actual gap in performance between recommended and non-recommended

products reported in Table V of -0.97% per year for equally-weighted portfolios.


29/51

$*

Evidence from value-weighted returns indicates that the performance of recommended

and non-recommended products as reported is very similar to the performance figures obtained

after eliminating the first one to three years of reported history (regardless of the model used to

measure performance). This suggests that backfill bias is not a problem among those products

which also report assets under management or, to put it differently, that those fund managers that

have backfilled data have probably backfilled only the return data.

A second concern is the impact of product size on returns. Tables I and III show that

investment consultants tend to concentrate on larger products. Previous research has shown that

funds that manage more assets perform worse (see Chen et al. (2004)), a finding that is consistent

with results in Table V showing that funds recommended by consultants perform worse (in

comparison with non-recommended products) on an equal-weighted than on a value-weighted

basis. We therefore re-assess the investment performance of recommended products in light of

this tendency. The choice of larger products may reflect the investment consultants optimization

of their own research and monitoring efforts, or a belief that recommending a large well-known

fund manager will be easier to justify after the event. However, the preference for recommending

large products may not be entirely free, for consultants may be required by plan sponsors to

recommend products of a certain size, perhaps because of doubts about the ability of small

products to handle a larger pool of assets.

To control for the impact of product size on performance, we use a regression-based

generalization of the calendar time portfolio approach (see Hoechle et al. (2009) and Dahlquist et

al. (2011)). This generalization relies on estimating, at the investment product level, a pooled

linear regression model with Driscoll and Kraay (1998) standard errors. Driscoll-Kraay standard

errors are robust to heteroskedasticity and general forms of cross-sectional and temporal


30/51

$+

dependence. The advantage of the regression-based approach is that it is straightforward to

include continuous and multivariate explanatory variables, and so to control for product size.

The pooled linear regression model we estimate has the following panel structure:

!!!! ! ! ! !!!!!!! ! ! ! !!

!!!!!

!

!! ! !!!! (6)

where ri,t is the excess return for product iin period t, and where we condition on time-varying

product characteristicszi,t . Product characteristics include investment consultant ratings, captured

by a full set of dummy variables, and de-trended log assets under management (AUM) at the end

of the previous period.25The risk-adjusted performance of recommended products is computed

using a four-factor model represented in equation (6) by the vectorft.

Table VIII provides the results of estimating different specifications of equation (6).

These specifications differ according to whether we control for lagged AUM, and whether we

include products with no information available on AUM. The coefficients of interest are those of

!p, since they inform us if recommendations and the other control variables have predictive

power over abnormal returns. In this table the constant captures the expected monthly excess

return or alpha of a non-recommended product (of average size in the models that include lagged

de-trended AUM as controls) and the coefficient associated with the recommended dummy

indicates the expected extra performance delivered by recommended products. Models I and IV

are the closest to the equal-weighted calendar time specifications showed in Table V. A

difference between the results in Table V and Models I and IV of Table VIII is that the panel we

25We use de-trended log AUM (de-trended by subtracting from the log of AUM the mean of this variable across all

period t observations) to address the possible non-stationarity of log AUM. However, almost identical results obtain

if we replace de-trended log AUM with standard log AUM.


31/51

%,

estimate in equation (6) is unbalanced and therefore the weight attached to each month in the

sample is not the same. The panel gives the same weight to each individual observation whereas

the portfolio method employed in Table V gives the same weight to each monthly observation.

Another source of difference between these results is that Table V shows results for

portfolios that include each product as many times as they have been recommended whereas

Table VIII does not. If we modify the weights in the panel regression to account for these

differences, as indicated in Hoechle et al. (2009), both results coincide.26 In model I, where

performance is assessed using industry-standard excess returns over benchmark indices,

recommended products underperform non-recommended products by 0.89% per year; less than

the 1.17% per year reported in Table V but still highly statistically significant. 27This difference

shrinks if we exclude from the sample products that do not report AUM (Model II) and

disappears when we control for the de-trended natural logarithm of AUM in each product at the

end of the previous year (Model III).

Similar conclusions can be drawn by looking at four-factor alpha models (Models IV to

VI), adjusted and unadjusted for product size. In model IV, recommended products underperform

non-recommended products by 0.38% per year but the difference is not statistically significant

this time. Moreover, once we control for lagged product size (Model VI), the underperformance

26Hoechle et al. (2009), following Loughran and Ritter (2000), argue however that weighting all observations

equally makes more sense.

27Annualized differences are computed by multiplying monthly coefficients times 12. Notice that since monthly

returns are linear in the (same) set of variables, it does not matter where the difference is calculated (i.e., for which

values of the lagged AUM, the variable used as control).


32/51

%#

of recommended products disappears or even turns into a small, but still not statistically

significant, outperformance (0.32% per year).

Finally, we look at the performance of portfolios of products that experience a net

increase (decrease) in the number of recommendations they receive. This analysis has two

objectives: first, to explore whether consultants failure to add value by their recommendations is

due to their inability to identify outperforming products or to the fact that they keep them on their

shortlist of recommended products for too long; and, second, to provide an alternative benchmark

in the performance analysis, by concentrating on products that are -./-01234.5678 3. 290 choice

set of (at least some of) the consultants we study.

We proceed by forming two different portfolios. The first portfolio includes all

investment products that experience a net increase in the number of recommendations received;

these are products that, in net terms, are being added to the shortlist of recommended products.

The second portfolio includes all investment products that experience a net decrease in the

number of recommendations received; these are products that, in net terms, are being dropped

from the shortlist of recommended products. Products are included in these portfolios in the

month in which they experience the increase/decrease in the number of recommendations and

kept there for 12 or 24 months. Each product is included in these portfolios as many times as it is

newly recommended/de-recommended, thus giving more weight to products that experience a

larger increase/decrease in the number of recommendations received.28

28It is possible that some of the portfolios that we consider to have experienced an increase/decrease in the number

of recommendations did in fact receive the same number of recommendations as in the previous year, and that we

capture instead the effect of the changing composition of the survey. However, because the coverage of the sample is

large and relatively stable through time, this problem, which may add noise to our estimates, is likely to be limited.


33/51

%$

Table IX compares the performance of these portfolios. Results indicate that products that

experience a net increase in the number of recommendations do no better than products that are

on average being de-recommended by consultants. In fact they do worse: the difference in

performance goes from a few basis points per year to more than three percent per year in some

specifications, but it is never statistically significant:290 number of products in each portfolio is

considerably smaller than in previous tables).

One of the objections that could be made against the results we presented in the previous

section is that some of the non-recommended products included in the analysis are effectively

off-limits to the consultants (because they are below the size threshold of plan sponsors, because

they lack sufficient longevity, or for some other reason). However, results in Table IX suggest

that consultants fail to identify superior future performers, even among the set of products that

are on their radar. We conclude that the underperformance of recommended products can be

explained by consultants tendency to recommend relatively large products. This may reflect the

fact plan sponsors and consultants are generally concerned about liquidity risk, and plan sponsors

avoid products that are small relative to their holding size.29At the same time investment

consultants may be inclined to focus on products that can be used broadly across their client base;

this would economize the costs of monitoring client fund holdings, and would guard against a

liquidity shortage in a product if, following a rating downgrade by the consultant, many plan

sponsors withdrew their assets simultaneously. However, even allowing for this constraint,

recommended products still fail consistently to outperform other products in our sample.

29We understand, for example, that plan sponsors would typically avoid holding more than 10% of the total assets

under management of a product.


34/51

%%

4. Conclusions

Using survey data from investment consultants with a combined share of around 90% of the

consulting market, we analyze their recommendations of U.S. active equity products over the

period 1999-2011. We examine the drivers, impact, and accuracy of these recommendations.

We find first that, while consultants recommendations of fund products are partly a

function of the past performance of those products (and of product size and fees), it is mainly the

fund managers non-performance attributes that drive recommendations. Consultants are not

merely return-chasing when they form their recommendations. Second, we find that investment

consultants recommendations have a large and significant effect on institutional asset allocation.

Third, we find no evidence that consultants recommendations add value to plan sponsors.

In addressing the reasons why consultants recommendations fail to add value, we find a

tendency of consultants to recommend large funds, which perform worse. There could also, of

course, be a simple lack of skill. However, it is also possible that better performing fund

managers are able to attract assets from plan sponsors without investment consultants

recommendations, whereas worse performing fund managers rely on consultants

recommendations to win such business. In this case worse performing fund managers would

make more effort to cultivate consultants than better performing managers. If this effort is

successful it would result in investment consultants recommending worse performers.

Our analysis shows performance in gross and net terms. The cost of pursuing a strategy of

picking actively-managed funds, encouraged and guided by investment consultants, is

considerable: the institutional funds in our sample charge, on average, 65 basis points a year,

which is far in excess of the cost of alternative index-related strategies. Moreover, plan sponsors

pursuing an active strategy tend to switch managers more often than those adopting an indexed


35/51

%&

approach, incurring transition costs which further widen the gap between the two approaches.

Consultants face a conflict of interest, as arguably they have a vested interest in complexity.

Proposing an active U.S. equity strategy, which involves more due diligence, complexity,

monitoring, switching, and therefore more consultancy work, drives up consulting revenues in

comparison to simple, cheap solutions. This is an important topic for further research.

Whatever the reasons for the lack of added value in consultants recommendations, it is

striking that fund sponsors follow such recommendations to the extent, and at the expense, that

they do. A possible explanation is that plan sponsors have reasons to engage investment

consultants other than their fund manager recommendations; these reasons could include the

hand-holding service described by Lakonishok et al. (1992), or the relationship of trust that

investors build up with their adviser (see Gennaioli et al. (2014)).

However, while these reasons might explain why plan sponsors engage investment

consultants in general, they do not tell us why they follow consultants recommendations when

they are apparently not rewarded for doing so. If plan sponsors know that they are not being

rewarded for following consultants recommendations, one possible reason for doing so is to hide

behind consultants recommendations when they have to account for their decisions. As Goyal

and Wahal (2008) find, plan sponsors that are more likely to be sensitive to adverse publicity

(headline risk) are more liable to use investment consultants. Jones and Martinez (2014) put

forward evidence that plan sponsors follow consultants recommendations even against their own

judgment. The tendency for plan sponsors to follow investment consultants, in the knowledge

that their recommendations do not add value, would be evidence of an agency problem.

It is, however, unlikely that plan sponsors can reliably judge whether investment

consultants add value or not. While fund managers testify to the rigor with which investment

consultants scrutinize their performance, investment consultants themselves are shy of disclosing


36/51

%'

the sort of information which would allow plan sponsors, or any outsider, to measure their own

performance. Among the consultants whose aggregate recommendations we have analyzed, some

will do better than others, and knowledge of differential performance would inform a plan

sponsors decision about which consultant to appoint. As it is, plan sponsors are making

appointments partly uninformed, and some may be nave about the actual utility of consultants

recommendations. An obvious policy response by regulators, or a market response by plan

sponsors, is to require full disclosure of consultants past recommendations so that such decisions

are better informed and, as a consequence, their assets more efficiently allocated.


37/51

%(

References

AICIO, 2013, Outsourced Chief Investment Office Survey, February, 45-57.

Altnkl, Oya, and Robert S. Hansen, 2009, On the information role of stock recommendation

revisions,Journal of Accounting and Economics48, 1736.

Barber, Brad, Reuven Lehavy, Maureen McNichols, and Brett Trueman, 2001, Can investors

profit from the prophets? Security analyst recommendations and stock returns,Journal of

Finance56, 531563.

Barber, Brad M., Terrance Odean, and Lu Zheng, 2005, Out of sight, out of mind: The effects of

expenses on mutual fund flows, The Journal of Business78, 20952120.

Bergstresser, Daniel, John Chalmers, and Peter Tufano, 2009, Assessing the costs and benefits of

brokers in the mutual fund industry,Review of Financial Studies22, 4129 4156.

Blake, Christopher and Matthew Morey, 2000, Morningstar ratings and mutual fund performance,

Journal of Financial and Quantitative Analysis35, 451483.

Busse, Jeffrey, Amit Goyal and Sunil Wahal, 2010, Performance and persistence in institutional

investment management,Journal of Finance65, 765790.

Cameron, A. Colin, and Pravin K. Trivedi, 1986, Econometric models based on count data.

Comparisons and applications of some estimators and tests, Journal of Applied

Econometrics1, 2953.

Carhart, Mark, 1997, On persistence in mutual fund performance,Journal of Finance52, 5782.

Chen, Joseph, Harrison Hong, Ming Huang and Jeffrey D. Kubik, 2004, Does fund size erode

mutual fund performance? The role of liquidity and organization, American Economic

Review94, 12761302.

Chevalier, Judith, and Glenn Ellison, 1997, Risk taking by mutual funds as a response to

incentives,Journal of Political Economy105, 11671200.

Coggin, T. Daniel, Frank Fabozzi, and Shafiqur Rahman, 1993, The investment performance of

U.S. equity pension fund managers: An empirical investigation, Journal of Finance 48,10391055.

Cremers, Martijn, Antti Petajisto, and Eric Zitzewitz, 2013, Should benchmark indices have

alpha? Revisiting performance evaluation, Critical Finance Review2, 1-48.

Dahlquist, Magnus, Jose Vicente Martinez and Paul Sderlind, 2011, Individual investor activity

and performance, Working paper.


38/51

%)

Del Guercio, Diane, and Paula Tkac, 2002, The determinants of the flow of funds of managed

portfolios: Mutual funds versus pension funds, Journal of Financial and Quantitative

Analysis37, 523557.

diBartolomeo, Dan, and Erik Witkowski, 1997, Mutual fund misclassification: Evidence based

on style analysis,Financial Analysts Journal53, 3243.

Driscoll, John C. and Aart C. Kraay, 1998, Consistent covariance matrix estimation with spatially

dependent panel data, The Review of Economics and Statistics80, 549560.

Fama, Eugene, and Kenneth French, 1993, Common risk factors in the returns on stocks and

bonds,Journal of Financial Economics33, 356.

Ferson, Wayne, and Kenneth Khang, 2002, Conditional performance measurement using

portfolio weights: Evidence for pension funds,Journal of Financial Economics65, 249

282.

Fung, William, and David A. Hsieh, 2000, Performance characteristics of hedge funds and

commodity funds: Natural vs. spurious biases, Journal of Financial and Quantitative

Analysis35, 291307.

Gennaioli, Nicola, Andrei Shleifer and Robert Vishny, 2014, Money doctors,Journal of Finance,

forthcoming.

Goyal, Amit, and Sunil Wahal, 2008, The selection and termination of investment management

firms by plan sponsors,Journal of Finance63, 18051847.

Hoechle, D., Schmid, M., Zimmermann, H., 2009, A generalization of the calendar time portfolio

approach and the performance of private investors. Working Paper University of Basel.

Inderst, Roman and Marco Ottaviani, 2012, Financial advice, Journal of Economic Literature50,

494-512.

Ippolito, R., 1992, Consumer reaction to measures of poor quality, Journal of Law and

Economics 35, 4570.

Jagannathan, Ravi, Alexey Malakhov, and Dmitry Novikov, 2010, Do hot hands persist among

hedge fund managers? An empirical evaluation,Journal of Finance65, 217255.

Jegadeesh, Narasimhan, Joonghyuk Kim, Susan Krische, and Charles Lee, 2004, Analyzing the

analysts: When do recommendations add value?Journal of Finance59, 10831124.

Jensen, Michael C., 1968, The performance of mutual funds in the period 1945-1964, Journal of

Finance23, 389416.


39/51

%*

Jones, Howard, and Jose Martinez, 2014, Institutional investor expectations, manager

performance, and fund flows, Working paper, available at SSRN 2252122.

Kelleher, Ellen, 2013, New York public pensions investigated,Financial Times, 10 November.

Khorana, Ajay and Edward Nelling, 1998, The determinants and predictive ability of mutual fund

ratings,Journal of Investing7, 6166.

Lakonishok, Josef, Andrei Shleifer, and Robert Vishny, 1992, The structure and performance of

the money management industry, Brookings Papers: Microeconomics, 339391.

Loughran, Tim, and Jay R. Ritter, 2000, Uniformly least powerful tests of market efficiency,

Journal of Financial Economics55, 361-389.

Mullahy, John, 1997, Instrumental-variable estimation of count data models: Applications to

models of cigarette smoking behaviour,Review of Economics and Statistics, 79, 586593.

Newey, Whitney K. and Kenneth D. West, 1987, A simple, positive semi-definite,

heteroskedasticity and autocorrelation consistent covariance matrix, Econometrica 55,

703708.

Pensions and Investments, 2011, Consultant Directory, November 28.

Petersen, Mitchell A., 2009, Estimating standard errors in finance panel data sets: Comparing

approaches,Review of Financial Studies22, 435480.

Rogers, W., 1993, Regression standard errors in clustered samples, Stata Technical Bulletin, 13,

1923.

Securities and Exchange Commission, 2005, Staff Report Concerning Examinations of Select

Pension Consultants, The Office of Compliance Inspections and Examinations, U.S.

Securities and Exchange Commission, May 16.

Sharpe, William F., 1964, Capital asset prices: A theory of market equilibrium under conditions

of risk,Journal of Finance19, 425442.

Siedle, Edward, 2013, Rhode Island State Pension Admits History Of 'Pay-To-Play' And SEC

Inquiry,Forbes Investing, 29 April.

Sirri, Erik, and Peter Tufano, 1998, Costly search and mutual fund flows, Journal of Finance53,

15891622.

White, H., 1980, A heteroskedasticity-consistent covariance matrix estimator and a direct test of

heteroskedasticity,Econometrica48, 817838.


40/51

%+

Womack, Kent L., 1996, Do brokerage analysts recommendations have investment value?

Journal of Finance51, 137167.

Wooldridge, Jeffrey M., 2010, Econometric Analysis of Cross Section and Panel Data, 2nd edn.

(MIT Press).


41/51

&,

Table I

Descriptive Statistics

The table presents descriptive statistics on the sample of investment consultants and institutional investment products used in our

study. Products are classified as recommended (Rec.) and not recommended (Not Rec.). Average asset size is in millions of US

dollars. The market cap-style based statistics in the second part of the table are averages over the 13 year period covered by the

sample (1999 to 2011).

Number of

Investment

Consultants

Number

of Recs.

Number of Products Average Product Asset Size

Rec. Not Rec. Total Rec. Not Rec. Total

1999 25 459 116 849 965 7,911 560 1,871

2000 36 1,398 241 856 1,097 5,624 737 2,140

2001 27 966 230 993 1,223 4,168 828 1,659

2002 32 1,434 314 1,266 1,580 2,757 632 1,150

2003 30 1,444 357 1,306 1,663 3,244 721 1,382

200430 1,745 409

1,913 2,322 4,056 1,079 1,709

2005 29 1,940 452 1,959 2,411 3,925 994 1,641

2006 28 2,107 503 1,930 2,433 4,198 984 1,733

2007 29 2,297 526 1,909 2,435 3,836 1,108 1,749

2008 30 2,164 557 1,842 2,399 2,611 650 1,138

2009 29 1,887 533 1,742 2,275 2,982 655 1,219

2010 27 1,608 476 1,672 2,148 3,481 798 1,414

2011 28 1,501 454 1,537 1,991 3,549 864 1,490

Large Cap Growth 29 437 90 345 435 6,045 927 2,185

Large Cap Value 29 455 91 315 406 5,610 1,257 2,334

Mid Cap Growth 24 150 38 121 159 2,216 513 958Mid Cap Value 25 108 29 92 121 2,753 564 1,169

Small Cap Growth 26 160 50 204 254 1,309 483 664

Small Cap Value 27 167 51 191 242 1,519 499 746

Core 23 316 104 491 595 3,147 902 1,332


42/51

Table II

Product Fees

This table shows average fees of the recommended (Rec.) and not recommended (Not Rec.) institutional investment products

used in our study. Panel A shows fees in the eVestment dataset as of the last year of our sample, 2011, or latest available. Panel B

shows average fees in the Informa Investment Solutions (IIS) dataset over the 13 year period covered by our study (1999 to 2011).Fees are in percent per year.

Panel A -

eVestment

Fees $10M Fees $50M Fees $100M

Rec. Not Rec. All Rec. Not Rec. All Rec. Not Rec. All

Large Cap

Growth0.67 0.72 0.71 0.60 0.63 0.62 0.54 0.57 0.56

Large Cap

Value0.67 0.70 0.69 0.57 0.60 0.59 0.51 0.55 0.54

Mid Cap

Growth0.77 0.80 0.79 0.73 0.71 0.72 0.68 0.66 0.67

Mid Cap

Value

0.79 0.82 0.81 0.71 0.72 0.71 0.63 0.66 0.65

Small Cap

Growth0.96 0.94 0.95 0.92 0.87 0.89 0.87 0.82 0.83

Small Cap

Value0.96 0.94 0.95 0.89 0.88 0.88 0.83 0.83 0.83

Core 0.63 0.72 0.70 0.57 0.62 0.61 0.52 0.57 0.56

All 0.75 0.78 0.77 0.68 0.69 0.69 0.63 0.63 0.63

Panel B - IISFees $10M Fees $50M Fees $100M

Rec. Not Rec. All Rec. Not Rec. All Rec. Not Rec. All

Large Cap

Growth0.71 0.77 0.76 0.61 0.62 0.61 0.54 0.56 0.55

Large CapValue

0.75 0.71 0.72 0.56 0.56 0.56 0.50 0.50 0.50

Mid Cap

Growth0.79 0.83 0.82 0.73 0.73 0.73 0.68 0.68 0.68

Mid Cap

Value0.83 0.83 0.83 0.71 0.69 0.70 0.63 0.62 0.63

Small CapGrowth

0.96 0.96 0.96 0.91 0.87 0.88 0.86 0.82 0.83

Small Cap

Value0.96 0.95 0.96 0.84 0.86 0.87 0.81 0.81 0.82

Core 0.65 0.74 0.72 0.54 0.58 0.57 0.52 0.52 0.52

All 0.81 0.82 0.82 0.68 0.68 0.68 0.62 0.63 0.63


43/51

&$

Table III

What Drives Consultants' Recommendations?

This table reports pooled time-series cross-sectional Poisson and Negative Binomial regressions of the number of investmentconsultants recommendations received by a product on past gross performance measures, asset management fees and variables

capturing soft investment and service characteristics of the asset managers as perceived by the consultants. Soft investment

factors, and service factors, are expressed using the fractional rank of each asset manager in the sample. An asset manager's

fractional rank, for a given variable, represents its percentile rank relative to other asset managers in the same period, and ranges

from 0 to 1. Past gross performance measures (excess returns over benchmarks and three factor alphas) and fees (as at the end ofthe sample period) are expressed using the fractional rank of each product in its investment category. All regressions include a

lagged measure of return volatility, lagged assets under management (in $ billions) and a full set of time dummies (not reported).

Each column represents a separate regression. z-scores based on standard errors clustered at the product level are included in

parentheses. The second part of the table displays model-implied average marginal effects and the bottom part shows the squared

correlation between observed recommendations and model-predicted recommendations. ***, **, * denote statistical significance

at 1%, 5% and 10% levels respectively.

I II III IV

(Poisson) (Poisson) (Poisson) (Poisson)

Soft Investment Factors (t) 2.37 2.40

(17.18)*** (17.25)***- Consistent Inv. Philosophy (t) 1.18 1.19

(8.28)*** (8.34)***- Clear Decision Making (t) 0.39 0.39

(2.91)*** (2.97)***

- Capable Inv. Professionals (t) 0.84 0.84(6.14)*** (6.10)***

Service Factors (t) 0.43 0.42

(3.49)*** (3.43)***- Relationship Manager (t) 0.26 0.26

(2.69)*** (2.66)***

- Useful Reports (t) 0.02 0.01

(0.19) (0.12)

- Presentation to Consultants (t) 0.32 0.33(2.71)*** (2.82)***

Past Performance Rank - Return (t) 0.23 0.21

(3.03)*** (2.94)***

Past Performance Rank - Alpha (t) 0.16 0.15(2.23)** (2.20)**

Fees (T) 0.48 0.48 0.46 0.46

(4.08)*** (4.07)*** (7.57)*** (7.53)***

Assets Under Management (t-1) 0.02 0.02 0.02 0.02(9.22)*** (8.83)*** (14.83)*** (14.20)***

Return Volatility (t-1) 1.04 0.76 1.03 0.77

(1.01) (0.73) (1.49) (1.11)

Average Marginal Effects

Soft Investment Factors (t) 5.78*** 5.84***

Documents

Picking Winners? Investment Consultants' Recommendations of Fund Managers